Modeling and Correcting Building Boundary in ICESat-2 Spaceborne Laser Altimeter Data Considering the Extended Laser Spot Effect

Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2)/Advanced Topographic Laser Altimeter System (ATLAS) can obtain nearly continuous profiles of ground targets. At present, fusing with other satellite-based data sources such as images is an important usage for ICESat-2 data. Some scenarios such as urban areas and applications such as stereo photogrammetry require registration features with a high geolocation accuracy when fusing. The building boundaries would be very appropriate features, as they not only have obvious geometric features in ICESat-2 data but also have textural features that are distinct in images. However, in ICESat-2 geolocated photons, the building boundary normally expands more than its actual boundary. This nonnegligible blurring phenomenon of geometric boundaries is caused by the extended laser spot of ICESat-2. This study theoretically and practically proposes a solution for these blurred building boundaries in ICESat-2 data. We first derive a theoretical model to describe the spatial convolution of laser spots (or called the extended laser spot effect) on building boundaries and then propose a method to estimate the exact building boundary points from ICESat-2 data. The signal model and boundary point correcting method are validated using typical 23 targets in four tracks of ICESat-2 in Hutt City, New Zealand, where local airborne lidar points with high density and accuracy are available. After correcting the horizontal offset of ICESat-2, the horizontal accuracy of the determined building boundary locations has a root-mean-square error (RMSE) of similar to 1 m, which is much better than that directly obtained by the signal photons from the ATL03 and ATL08 product (with RMSEs of similar to 6 m). The results indicate that this study can provide significant feature points for accurate registration and fusion between ICESat-2 data and other data sources in urban areas.